RFC: Drop CUDA support on ppc64le

As a consequence of a lot of work in the CUDA 12.x stack by the nvidia folks leading up to #7431 & #7476, upon closure of the CUDA 12.9 migration we'd be able to fully decouple `cuda_compiler_version` from the GCC pins https://github.com/conda-forge/conda-forge-pinning-feedstock/blob/7b5f2c5cb1b357ec4c5c0ddeb464b6cfc4b907dd/recipe/conda_build_config.yaml#L152-L158 (c.f. #6967) The fly in the ointment is ppc64le, which CUDA dropped support for in 12.5. As such, we currently have https://github.com/conda-forge/conda-forge-pinning-feedstock/blob/7b5f2c5cb1b357ec4c5c0ddeb464b6cfc4b907dd/recipe/conda_build_config.yaml#L60-L63 which also forces us to keep GCC in the zip, because CUDA 12.4 doesn't support anything newer than GCC 12. Given that upstream CUDA has dropped ppc support, it's pretty much inevitable that we eventually drop support as well. We do provide a way (#7472) to re-add CUDA 11.8, which can be used to re-add ppc support on a per-feedstock basis. We could do the same for CUDA 12.4 if we wanted. Thoughts? @conda-forge/core @conda-forge/cuda